Road edge detection method

Abstract

The invention discloses a road edge detection method. The method comprises the following steps of S1, collecting road video images and preprocessing the images; S2, using a Gabor filter to extract a texture direction; S3, through a voting mode, determining an initial disappearance point position; S4, carrying out initial positioning on a road edge; and S5, carrying out disappearance point iteration re-estimation, using an iteration mode to continuously update a disappearance point and a road edge till that the distance between the two updated disappearance points at a previous moment and a later moment is less than threshold, and acquiring the correct road edge. In the invention, the Gabor filter is used to estimate the pixel texture direction, and simultaneously, a corresponding degree ofconfidence is calculated; according to the texture direction, a pixel with the degree of confidence which is greater than the threshold is used to vote and determine the initial disappearance point;the initial disappearance point is taken as a starting point, one group of rays are established downwardly, the direction consistency rate of color difference and double angle sine function weightingis calculated, and the ray corresponding to a maximum value is taken as an initial road edge; two side edges are alternatively sampled, the optimal disappearance point is selected, and two road edgesare acquired.

Description

technical field [0001] The invention specifically relates to a road edge detection method. Background technique [0002] In recent years, most of the research on road edge detection technology at home and abroad has focused on structured roads. Although good research results have been achieved, in practical applications, roads without obvious signs, such as rural roads and desert roads, are often encountered. Wait. Therefore, it is of great significance to find an unstructured road edge detection algorithm with strong anti-interference and high precision. [0003] For unstructured road detection, traditional methods based on road features, such as color, block size and texture, are easily affected by shadows and lighting. The method based on the road model not only has a large amount of calculation, but also has low detection accuracy and is easily disturbed by noise. These traditional detection methods are subject to their own constraints, resulting in a limited range of...

AI Technical Summary

Problems solved by technology

Although these improvements reduce the estimation error of the vanishing point, they all use the initial vanishing point as a constraint to detect the road edge, which easily leads to edge detection failure.